نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
Accurate prediction of actual evapotranspiration in response to water resource management problems and increasing the productivity of agricultural products under dry and semi-arid climatic conditions is considered essential. Sunflower, as an economic crop with relative resistance to water deficiency, is a suitable option for cultivation in areas with water limitations. In this study, Support Vector Regression (SVR) and Random Forest (RF) models optimized with the Genetic Algorithm (GA) were used to predict actual evapotranspiration. For this purpose, data from weighing lysimeters in Bushland, Texas, USA, during the 2011–2012 growing season were extracted. A total of 230 data points were selected as model inputs in the form of eight scenarios with different combinations of climatic parameters. The results of this study showed that, in the SVR-GA model, the best performance was obtained for the combination of inputs including daily soil temperature inside and outside the lysimeter at a depth of 1.6 m, relative water storage at midnight, daily average relative water storage, soil temperature at different depths, air temperature, soil heat flux in the rows and the distance between rows, relative humidity, and wind speed, such that the Root Mean Square Error (RMSE) was 1.017 mm/day, the correlation coefficient (R) was 0.95, and the Nash Sutcliffe Efficiently (NS) was 0.908. Also, in the RF-GA model, the best results were obtained with input variables including all the above-mentioned variables except wind speed, with an RMSE of 1.516 mm/day, an R value of 0.90, and an NS value of 0.796. Comparison of the results showed that the use of the Genetic Algorithm in optimizing the models increased prediction accuracy, such that based on the RMSE criterion, the hybrid SVR-GA and RF-GA models experienced reductions of 59.91% and 35.8%, respectively, compared with their corresponding standalone models. Finally, the hybrid models can be used as efficient tools for estimating actual evapotranspiration under conditions where direct measurement using traditional methods is costly or limited.
کلیدواژهها English
3- Bayat, V.M., Zare, A.H., Maroufi, S., Sabziparvar, A.A. and Soltani, F., 2009. Simulation of reference evapotranspiration using artificial neural method and empirical methods and comparison with experimental lysimeter data in cold semi-arid climate of hamedan.
